On the economics of malaria suppression

Emiel Sanders

April 2009

Supervisor:

Yvonne Adema

Master thesis International Economics and Business Studies

Erasmus University Rotterdam

2

To Mickey

3

Abstract

Using WHO-CHOICE methodology, we found several established malaria

interventions and the introduction of a vaccine and combinations thereof to be highly

cost-effective (8-127 int$/DALY) in sub-Saharan Africa, where the malaria disease

burden is highest. By means of Solow modeling, we estimate that continuous malaria

suppression increases per capita income up to 50% compared to a non-intervention

scenario after a time span of 65 years.

4

1. Introduction ...............................................................................................................5 2. Review of literature....................................................................................................5

2.1 Cost-effectiveness studies....................................................................................6 2.2 Macroeconomic studies .......................................................................................7

3. Methods....................................................................................................................12 3.1 Cost-effectiveness analysis malaria interventions .............................................14 3.2 Macroeconomic impact analysis malaria interventions.....................................17

4. Results ......................................................................................................................26 4.1 Cost-effectiveness study ....................................................................................26 4.2 Macroeconomic analysis....................................................................................29

5. Discussion................................................................................................................33 6. Conclusion ...............................................................................................................35

References ....................................................................................................................36 Appendix A...................................................................................................................39 Appendix B...................................................................................................................40 Appendix C...................................................................................................................41 Appendix D ..................................................................................................................47

5

1. Introduction

Approximately 1 million people die yearly of the consequences of malaria and more

than 3 billion individuals, spread over 107 countries in which malaria is endemic, are

at risk of contracting the parasitic disease (WHO (2005), Breman et al. (2004)). The

total estimated incidence amounted to 402 million cases in 2004 (Korenromp, 2004).

According to the WHO (2002), malaria is the eighth largest contributor to the global

burden of disease and the second largest to the disease burden in Africa in terms of

disability adjusted life years (DALY’s). In this continent 60% of total malaria

morbidity and over 90% of malaria related death occur. Malaria P. falciparum, one of

the four species that cause malarial infections in humans, is largely responsible for

this burden. 82% of total malaria related death comes about before the age of 5 and 5-

8% of child mortality in Africa is ascribed to malaria (Bryce et al., 2005). As effective

and affordable malaria control interventions exist, this human tragedy and the large

economic cost (Chima et al., 2003) can be mitigated. Scaling up existing interventions

and the introduction of new ones would reduce this toll and also contributes to

achieving the Millennium development goals (MDG’s), especially numbers 4 and 6,

i.e. the reduction of child mortality and the combat against malaria and other

diseases1. To do so, cost-effectiveness analysis is a tool to prioritize the gains and the

costs of interventions. We conduct such a study and estimate the efficiency of existing

methods and of the introduction of a malaria vaccine. In addition, we assess their

impact on macroeconomic growth. We estimate the direct and indirect effects of

malaria interventions on economic wellbeing.

The outline is as follows. We start our study with the review of literature that is

followed by a description of the methods used in our analysis. The results are

presented afterwards. At the end the discussion and the conclusion are found.

2. Review of literature

Below a short review of the literature on cost-effectiveness and macroeconomic

studies related to malaria are shown. At the end the value added of this study is

described.

1 http://www.un.org/millenniumgoals/

6

2.1 Cost-effectiveness studies

Many cost-effectiveness studies follow the methodology set by the WHO-CHOICE

project, an initiative installed by the WHO in 1998 to assist policy makers in their

decisions on interventions and programmes that maximize health gains, in terms of

disability adjusted life year (DALY), given budget constraints. An overview of

several control interventions is found on the WHO-CHOICE website

(http://www.who.int/choice/en/) and in a special issue of the British Medical Journal,

November 2005. However, this methodology has not always been the norm.

Considering malaria, Goodman et al. (2000) put forward an overview in which 15

other cost-effectiveness studies that use different methodologies are displayed. These

studies express effectiveness either in terms of morbidity or mortality. Those analyses

that use a combined measurement of effectiveness mortality and morbidity captured

by DALY and follow WHO-CHOICE’s guidelines are outnumbered.

Malaria interventions are classified in two groups. Those that reduce malaria

incidence are considered to be of a preventative nature and those that negatively affect

malaria mortality are the so-called curative interventions.

Using data on the burden of malaria (Najera et al., 1996 and Murray et al.,

1996) and the methodology as used in the Global Burden of Disease (Murray et al.,

1996), Goodman, Coleman and Mills (2000) find malaria control interventions to be

highly cost-effective. For insecticide treated bed nets (ITN) and indoor residual

spraying (IRS) they find respective costs (1995 US$) per DALY to range between 19-

85 and 16-29, respectively. Intermittent preventative treatment (IPT) for pregnant

women is estimated between 4-29 US$ per DALY. Curative interventions were also

analyzed and considered highly cost-effective. Chemoprophylaxis for children, and

improvements in case management, such as improving compliance and availability of

drugs, were estimated at a respective cost-effectiveness of 3-12 and 1-8 US$ per

DALY. Although these results are encouraging several remarks have to be made.

Firstly, the analysis is conducted for sub-Saharan Africa as a whole. Differences in

costing structures or malaria epidemiology are accounted for in a rough manner.

Secondly, it is unclear to what extent effectiveness of the malaria interventions,

obtained from the literature and control-trials, is discounted for factors that obstruct

them. Lastly, only individual interventions are compared. Intervention packages are

not scrutinized.

7

By means of a 10 year implementation program, Morel et al. (2005) estimated

the cost-effectiveness of seven malaria interventions and combination thereof for

different coverage levels in two sub-Saharan African regions in terms of 2000 US $

per DALY. Epidemiological data came from the 2002 Global burden of disease

(GBD) estimates. Data on efficacy and costs of the interventions were obtained from

experts and literature. Morel et al. (2005) calculated the net effectiveness of the

interventions given a baseline efficacy that is corrected for patients’ behaviour

(adherence to regimen), pharmacokinetics (success of intervention when regimen is

not followed) and resistance to drugs (biogenetics), regardless of geography. Total

costs are estimated by means of the CostIt model (Baltussen et al., 2003) and cover

unit costs, distribution costs, media costs and labour costs. The evaluated

interventions are case management with chloroquine (CQ), sulfadoxine-

pyrimethamine (SP), non-artemisinin based combination therapy (non-ACT) and

artemisinin based combination (ACT). Also insecticide treated nets (ITN), indoor

residual spraying (IRS), intermittent presumptive treatment (IPT) during pregnancy

were evaluated. The authors found all interventions except that of IPT to be highly

cost-effective (9-35 int$/DALY) at a 80% coverage level, however most rewarding

are high-coverage levels of ACT.

In their cost-effectiveness study, Breman et al. (2006) examine the following

interventions: ITN, IRS, IPT during pregnancy and a change of first line drugs. They

focus on the area of sub-Saharan Africa where malaria is highly endemic but stable.

Epidemiological data are obtained from Breman et al. (2004) and WHO (2002). Data

on costs and effects of interventions are obtained from different sources. The methods

used to perform the cost-effectiveness analysis are unclear. Again all interventions

evaluated (ITN, IRS and IPT during pregnancy) are considered to be highly cost-

effective as they range between 11-24 mean cost (in US$ 2001) per DALY.

2.2 Macroeconomic studies

Health has in general a positive and significant effect on economic growth. In terms

of mortality, Bloom et al. (2004) show in their overview that in 12 out of 13 studies

health has a positive and significant impact on economic growth. All studies use

cross-country regression analysis and employ other explanatory variables to estimate

the economic impact of health. In terms of morbidity, Mills and Shillcutt (2004)

demonstrate that this dimension of health has a major impact on economic activity

8

through its adverse effects on productivity, investment and education. When

considering poor health states, among that malaria, Cole and Neumayer (2006) argue

that morbidity plays a more important role than mortality in relation to economic

development.

Before turning to malaria, we highlight HIV/AIDS as the other communicable

disease that has been scrutinized in an macroeconomic context. The literature on the

economic modelling of HIV/AIDS can be sub-divided in three groups. The bulk of the

literature estimates the macroeconomic cost of HIV/AIDS by means of (augmented)

Solow models, using either a one-sector or two-sector neoclassical growth model,

specified to individual countries or in a cross-country manner. Among these works are

that of Over (1992), Cuddington (1993), Cuddington and Hancock (1992, 1994),

MacFarlan and Sgherri (2001) and Haacker (2002a, 2002b). All estimate the endemic

to have a negative effect on welfare, be it rather small as the full impact of the

epidemic has not yet been recognized at that time of the early writings. Haacker puts

forward frameworks for macroeconomic modelling of HIV/AIDS. These models are

popular among macroeconomists studying HIV/AIDS as in hardest hit areas data are

scarce and of poor quality. Moreover, the Solow model is considered the workhorse

model health impacts. The second group of macroeconomists uses CGE-techniques

for estimating the welfare impact of HIV/AIDS. Among those is Arndt and Lewis

(2001), that estimate the South African economy to slow yearly by 2.6 percentage

points over the 1997-2010 simulation period using a no-AIDS and an AIDS scenario.

The last group uses different econometric techniques to model the macroeconomic

impact of HIV/AIDS. Bloom and Mahal (1997) take into account effects of

simultaneity and possible correlations of factors that influence growth to determine

the effect of AIDS on economic well-being. Interestingly, their outcome estimates

that the epidemic has an insignificant effect on the growth rate of per capita income.

A study by Bonnel (2000) estimates that HIV/AIDS reduced economic growth over

the 1990-1997 period by 0.7% per year using cross-country regressions in Southern

Africa. The slowdown is exacerbated in the presence of malaria by 0.4% per year.

Many microeconomic studies on the impact of malaria on households and

firms have been conducted which are nicely summed up by Chima et al. (2003). The

amount of their macroeconomic counterpart is small. Those that have been performed

can be divided in three categories. The first group encompasses those studies

(Shephard et al.(1991), Ettling et al. (1991) and Leigthon et al.(1993)) that estimate

9

the economic burden of malaria on a macroeconomic level by summing up the direct

and indirect costs of malaria. The direct costs are the sum of the household and

government expenditures on malaria treatment and prevention. The indirect costs are

the losses of economic output due to mortality, morbidity and debility. These are

proxied by the adult output per day times the estimated productive time lost due to

both adult and childhood malaria episodes. These authors estimate a negative impact

of malaria on economic development of 0.6% and over of total GDP in several

African countries. Goodman et al. (2000) argue that the methodology of summing up

direct and indirect costs should be used with care as deviations in the base values have

strong effects on aggregate values.

Among the second group are the studies of Gallup and Sachs (1998) and

McCarthy et al. (2000) which estimated the impact of malaria on economic well-

being by means of regression analysis similar to the methodology of Barro (1991).

Gallup and Sachs (1998) conclude that for the 1965-1990 period, countries with

severe malaria had much lower economic growth amounting to 1.3% per year in terms

of GDP per capita. In their regression they include several factors likely to influence

economic growth. These are initial income levels, initial human capital stock

(measured by secondary school enrolment rates and life expectancy at birth),

economic policy (measured by trade openness and an index of the quality of public

institutions) and geographical variables (measured by an indicator for the

geographical tropics and the fraction of the population within 100 km of the coast). In

their study malaria is approximated by the malaria index. This is the product of the

percentage of malaria endemic land area in 1965 and the percentage of malaria P.

falciparum cases in 1990. In addition, they conclude that a 10% reduction in the

malaria index would have resulted in a 0.3% increase in terms of GDP per capita per

annum over the same period.

McCarthy et al. (2000) assess the economic impact of malaria in terms of

average per capita GDP by using several explanatory variables. These are the

investment ratio, primary and secondary school enrolment rates, initial income per

capita, openness measured as total trade per total GDP, political freedom, the number

of revolutions per year, an index of assassinations and a malaria morbidity variable.

The latter is proxied by the incidence of malaria episodes over three five-year periods

starting in 1983 complemented by data on sanitation, climate and location, health

expenditures, among others. By including these variables the effectiveness of the use

10

of protective malaria measurements across regions is taken into account. Gallup and

Sachs’ (1998) malaria variable lacks such a dimension. An average over five years is

taken to overcome misrepresentation as malaria incidence varies strongly over time.

The authors find a strong and significant negative impact of malaria on economic

development, however the impact differs sharply among countries. For the most

countries annual per capita growth is reduced by a quarter percent while an average of

-0.55% GDP growth is estimated for sub-Saharan countries.

The differences between the findings of Gallup and Sachs (1998) and

McCarthy et al. (2000) are partly assigned to the different periods of analysis. The

former authors analyze the economic impact of malaria for the 1965-1990 period,

while the latter consider the 1983-1998 period. Mills and Shillcutt (2004) argue that

most progress in reducing the adverse consequences of malaria has been

accomplished prior to the timeframe used by McCarthy et al. (2000). In addition, the

malaria variable used by McCarthy et al. takes into account several other explanatory

variables which they think nuance the adverse economic effects of malaria. The

malaria variable used by Gallup and Sachs is primarily focused on geography and

prevalence. Also the use of a different set of control variables may contribute to the

different outcomes of both studies.

The third group of macroeconomic studies on malaria encompasses the study

of Barlow (1967). In his article Barlow uses a Cobb-Douglas production function with

skilled and unskilled labour and capital as input factors to estimate the impact of

malaria eradication on long-run GDP per capita in Sri Lanka (Ceylon). Labour is

disaggregated into skilled and unskilled labour as it is assumed that malaria is a low

income disease mainly because it is avoidable. To estimate this impact, two scenarios

on the basis of demographic differentials of eradication and non-eradication are run

and compared to each other. It is assumed that eradication has a direct effect on labour

inputs through its effects on mortality and fertility and on morbidity and debility.

Indirect effects on capital inputs occur through changing public and private saving

rates. Barlow estimates that in the short run GDP per capita is positively affected up

to 10% due to malaria eradication. In the long run, however, population growth

diminishes this gain. Several authors questioned Barlow’s study. With respect to our

study, Frederiksen (1966), Borts (1967) and Brown (1986) are most important. They

argue that Barlow failed to take into account the economic effect of malaria

11

eradication through increased agricultural productivity in areas that were previously

impassable due to malaria endemicity.

In this study a similar model to that of Barlow (1967) is used, but it differs in

several ways. Firstly, we do not disaggregate the labour population into skilled and

non-skilled as according to Gallup and Sachs (1998) malaria does not differentiate

between rich and poor people. Secondly, we take into account the indirect effects of

(partial) malaria eradication through its effect on TFP. The elasticity of TFP to

malaria incidence estimated by Cole and Neumayer (2006) is used for this purpose.

By including the indirect effects, we take into account the argument made by Gallup

and Sachs (1998) who state that these effects are substantial. Thirdly, contrary to

Barlow who argues the malaria intervention costs are insignificant, we do include

them to estimate the net effects of the proposed malaria interventions on economic

well-being. Lastly, unlike Barlow’s model that disaggregates savings into private,

government and foreign savings, we consider the total savings rate.

Criticism made by Frederiksen (1966) and others on the positive effects of

malaria eradication on land use is acknowledged in the literature study by Goodman

et al. (2000). However, the relationship has not been academically validated and

therefore not been quantified. Only one study by Wang’ombe and Mwabu (1993) is

present that examines the relationship between malaria and land use. They find at the

household level insignificant effects of malaria on the cassava production and

cultivated land area, which they attribute to the coping strategies when households are

affected by malaria. Goodman et al. (2000) point to methodological weaknesses in the

Wang’ombe et al. (1993) study. They state that the impact of malaria on long-term

land exploitation cannot be derived from household studies. So, the impact of malaria

on land use is acknowledged by several authors, however the negative relationship

has not been proved. Therefore, we abstain from including land as a distinctive

production factor of GDP and focus solely on TFP, total labour and capital stock.

Nevertheless, Frederiksen et al. point to an anomaly which has also been mentioned

by Gallup and Sachs (1998). Due to malaria, investments in profitable sectors such as

tourism (and agriculture) are abstained from. Eradication, therefore, would increase

the return on investment in these sectors and attract investors. The Solow model does

not allow for these behavioural changes. A change in the savings rate could

circumvent this stringency of the Solow model, however we prefer maintaining a

constant savings rate for two reasons. Firstly, few studies on the savings behaviour in

12

relation to malaria, both on a micro and a macro level, exist to fully capture this

conduct (Chima et al.,2003). Secondly, based on Modigliani’s life cycle theory,

adjusting household saving rates upward would pose intergenerational consumption

problems. We think the current generation is unwilling to increase savings as this

sacrifice in lower current consumption is too large. Moreover, when increasing

government savings it is likely that the government is unable to balance its budget as

the sum of total government consumption and investments increase total taxes. By

adjusting TFP (and therefore GDP) for malaria incidence, we sidestep the inability of

the Solow model to anticipate on higher returns on investment when the malaria

burden is brought down.

With our study we contribute to the overall literature in several ways. Firstly, the

estimation of the cost-effectiveness of a malaria vaccine is a novelty. Secondly, the

value added is to be found in introducing a model which directly links the effects of

health interventions on economic growth through the stock of labour. Thirdly, to our

knowledge, this study is the first in estimating the economic impact of malaria

interventions for Africa on a macroeconomic level and in assessing which of these are

most efficient.2 Lastly, our model enables us to quantify the direct and indirect effects

of malaria interventions separately in contrast to the studies using a Barro (1991)

regression methodology.

3. Methods

Below the methodologies of the cost-effectiveness analysis of malaria interventions as

well as the macroeconomic impact study are described. Both studies cover 4 regions

located in sub-Saharan Africa (as defined by the MNP-institute) in which P.

falciparum is highly endemic. Table 1 gives a list of countries by region.

2 On a macroeconomic level Abegunde et al. (2007) estimated the cost of chronic diseases by means of a Solow model. Their methodology is different from ours as they only include the direct effects of reducing the prevalence of chronic diseases. Further studies on health interventions on macroeconomic level are to our knowledge non-existent.

13

Region 1 Western Africa

Region 2 Central Africa

Cape Verde Chad Benin Gambia Ghana Guinea Côte d'Ivoire Liberia Mali Mauritania Niger Nigeria Guinea-Bissau Saint Helena Sao Tome and Principe Senegal Sierra Leone Togo Burkina Faso

Cameroon Central African Republic Congo Congo, the Democratic Republic of the Equatorial Guinea Gabon

Region 3 Eastern Africa

Region 4 Southern Africa

Burundi Comoros Ethiopia Eritrea Djibouti Kenya Madagascar Mauritius Réunion Rwanda Seychelles Somalia Sudan Uganda

Angola Botswana Lesotho Malawi Mozambique Namibia Zimbabwe Swaziland Tanzania, United Republic Zambia

Table 1 Countries by region.

We focus on these regions due to its high malaria mortality (over 90% of world

fatalities (WHO mortality database, 2005)) and malaria prevalence (61% of world

total, (Korenromp, 2005)). C++ implementation in M programming language

(http://www.my-m.eu/) is used to perform the analyses. All monetary values are

expressed in international dollars, int$, i.e. US dollars for the year 2000. Data have

been converted to region specific values, when necessary by means of a weighted

average.

14

3.1 Cost-effectiveness analysis malaria interventions

We determine the cost-effectiveness of malaria control interventions by estimating

their total costs and their influence on population health using a simple demographic

model. In doing so, we follow the guidelines as formulated by the WHO-CHOICE

project (http://www.who.int/choice/en/), which enables us to comprehensively assess

the health impact of the interventions. A gradual approach of this methodology is

shown in appendix A, which is derived from Niessen et al. (2009).

Demographics and malaria epidemiology

Based on the 2005 population stock across region, sex and 1-year age cohorts

(aged between 0-100) obtained from the MNP-institute, we annually simulate

population dynamics by means of fertility and mortality rates.

Constant total fertility rates (MNP, 2000) and a constant birth-ratio are used to

estimate total birth across region and sex. We consider women of child bearing age

(WCBA) to give birth when aged between 15 and 50 years old. The birth-ratio, ratio

of newborn boys over girls, is estimated on the basis of the 2005 population stock.

Total birth is kept constant during the cost-effectiveness analysis.

Total death across sex, age and region is composed of malaria and non-malaria

related death. Total and malaria mortality rates over sex and age (0-1, 1-4, 5-year age

cohorts between age 5 and 85 and 85+) are derived from the WHO 2005 mortality

database. Background mortality is calculated as the differential of the two rates, which

we keep constant over time.

Modelling is such that a time lag of one year is present before total birth enters

population dynamics. Death is modelled instantaneously.

The malaria epidemiology is derived from the core demographic model as

described above. Malaria prevalence rates are calculated as the product of incidence

rates across region, sex and age (0-4, 5-14, 15+), a two-week average per malaria

episode and the total population stock that is estimated by the demographic model.

Incidence rates are obtained from Korenromp (2005). Malaria related death is

calculated by the product of total malaria mortality rates and total population stock.

Table 2 shows total population and an estimation of the burden of malaria across

regions in 2005.

15

Total population Incidence Gross incidence rate Deaths DALY's*3

Region 1 276,011,300 117,264,000 0.42 490,801 15,406,410

Region 2 83,785,560 34,275,940 0.41 154,549 4,555,666

Region 3 225,042,700 46,745,320 0.21 165,437 5,787,988

Region 4 120,388,300 43,949,840 0.37 106,961 3,417,455

Total 705,227,860 242,235,100 0.34 917,748 29,167,519

Table 2 Total population stock and burden of malaria across region, 2005. * Disability adjusted life years.

Malaria interventions

Few means are available to combat malaria. To simplify our analysis we evaluate the

cost-effectiveness of five malaria control interventions and combinations thereof

instead of the seven interventions as put forward by Morel et al. (2005). We abstain

from evaluating several curative methods, which are SP, non-ACT and IPT during

pregnancy. However, we evaluate the introduction of a vaccine. So, we focus on three

preventative interventions (ITN, IRS and a vaccine) and two curative interventions

(case management with CQ and ACT) and combinations thereof. In total we analyze

19 intervention packages. Although SP is more effective than CQ and both have the

same cost-basis (Morel et al., 2005), we focus on the latter as it is used in all focus

regions for combating severe malaria P. falciparum (World malaria report, 2008). We

refrain from evaluating SP during pregnancy due its low effectiveness and high cost

(Morel et al., 2005). We choose ACT instead of non-ACT because the former is

adopted as the first line treatment for severe P. falciparum in all focus regions. In

addition, non-ACT is less effective than ACT and costs approximately the same as

ACT (Morel et al., 2005, World malaria report 2008).

We use the data of Morel et al. (2005) on the net effectiveness and the costs of

all interventions, excluding the vaccine to determine their cost-effectiveness. Table 3

shows the net effectiveness of the individual malaria control interventions.

Intervention Reduction in incidence (%) Reduction in case fatality (%)

Insecticide treated bed nets (ITN) 50 20

Indoor residual spraying (IRS) 50 20

Vaccine 65.9 26.4*

Case management CQ 27

Case management ACT 63

Table 3 Net effectiveness individual malaria interventions. * same proportional reduction in incidence and case fatality as used for ITN and IRS is applied.

We employ the conclusions of Aponte et al. (2007) on the net efficacy of a candidate

malaria candidate vaccine, which they tested in a highly endemic area of

3 The burden of malaria is estimated by means of the WHO-CHOICE methodology by eliminating malaria related morbidity and mortality.

16

Mozambique. The total costs of the vaccine is composed of two parts, the vaccine and

the non-vaccine costs. The former we obtained from expert’s opinion and the latter is

a conservative approximation of pneumococcal non-vaccine costs estimated by Sinha

et al. (2007). See appendix B for an overview of the evaluated interventions and their

costs.

The effects of these interventions and their combinations on the region

specific malaria incidence and mortality rates enable us to estimate the regional

burden of disease. By doing so, we adjust regionally for the coverage level of

established health care facilities that hamper malaria transmission and the number of

malaria fatalities (World malaria report 2005, 2008). Table 4 displays the coverage

levels of the selected malaria control interventions.

Intervention Region 1 Region 2 Region 3 Region 4

Insecticide treated bed nets 2.6% 0.9% 3.0% 2.9%

Indoor residual spraying 0.3% 0% 4.4% 8.0%

Vaccine 0% 0% 0% 0%

Case management CQ 38.7% 44.4% 14.2% 36.6%

Case management ACT 5.6% 1.6% 18.8% 41.7%

Table 4 Coverage levels malaria control interventions across region.

In line with the WHO-CHOICE methodology, we assume the interventions to be

effective for a period of 10 years after which malaria incidence and mortality rates

return to pre-intervention levels. All interventions but the vaccine apply to the 80% of

the full population, i.e. we increase coverage levels to 80%. A conservative approach

for implementing a malaria vaccination programme is assumed: only newborns are

vaccinated. Evaluation is performed 100 years after the intervention programme ends

to include all healthy life years gained. We assume that resistance to the curative

methodologies does not occur. Model uncertainties are accounted for by means of

hazard equations which assume a normal distribution when calculating risks of

contracting malaria and dying of the consequences of it.

Health gains are expressed as the average number of disability adjusted life

years (DALY’s) averted. These are calculated by applying region-specific disability

weights for the general population by age and sex

(http://www.who.int/choice/demography/health_valuations/en/index.html). When

infected with malaria, a universal disability weight of 0.6 is assigned (GBD, 1990).

Costs are calculated in yearly averages. Both health gains and costs are discounted

17

over time by a rate of 1.5% and 4.5%, respectively. Then, the division of costs over

health gains gives us the desired cost-effectiveness ratio in terms of int$/DALY.

To determine to what extent the evaluated interventions contribute to the Millennium

Development Goals (MDG’s) we estimate their impact on under 5 mortality. We do

so by taking a conservative approach and analyze the impact of the most cost-

effective interventions as we believe policymakers will choose those interventions

that are most efficient. Again during 10 years interventions are assumed to be

effective after which malaria incidence and mortality return to pre-intervention levels.

Once more, only newborns are vaccinated. The other interventions apply to the whole

under 5 population. Then, the average proportional difference in malaria mortality

summed over sex and under 5 population over 15 years is taken, which will be our

impact measurement of the interventions on malaria mortality.

Lastly, we analyze if eradication of malaria mortality is feasible given the

effectiveness of the malaria control interventions. We do so by scrutinizing their

impact on malaria mortality rates.

3.2 Macroeconomic impact analysis malaria interventions

To determine the macroeconomic impact of the malaria packages as described in

appendix B we link the demographics and epidemiology model applied in the cost-

effectiveness analysis with the general Solow model for closed economies. We

calculate the direct effects of the interventions and complement these results by

accounting for the indirect effects of malaria suppression and compare both with a

non-intervention scenario. When the labour force is fully covered by the

interventions, in 2070, we evaluate the impact on economic well-being in terms of

GDP per capita.

Four channels are present through which we project the effects of malaria

interventions on economic well-being. Firstly, malaria mitigating policies negatively

influence malaria incidence and mortality. Therefore, when intervening, total labour

stock expands. By means of our demographics and epidemiology model we project

these effects of interventions on total labour stock over time and across regions.

Secondly, when infected with malaria one is confronted with a productivity loss. We

18

account for this loss through the disability weight of malaria. Successful interventions

reduce this loss and therefore the labour force becomes more productive. Thus, when

effectively intervening in malaria prevalence the labour stock increases in size and

strength. These two dimensions are captured by the effective labour variable

(discussed below) that is a production factor in the Solow model. Thirdly, if the

intervention costs are borne by the economy under consideration, savings are diverted

away from investments and total GDP is negatively influenced as capital

accumulation is hampered. Through these three channels (size and vigour of the

labour force and capital accumulation) economic well-being is directly influenced.

Indirectly, malaria tends to affect economic growth through its undesirable effects on

total factor productivity (TFP), as pointed upon by Gallup, Sachs and Mellinger

(1998) and Cole and Neumayer (2006). Due to malaria endemicity, investments are

refrained from and the transmission of ideas and knowledge is hindered through

limitations of internal movement. By adjusting for these effects on TFP, we strive to

take into account their arguments and estimate the indirect effects of malaria

suppression.

3.2.1 Demographics and epidemiology model

We build upon the demographics and epidemiology model as described in the

previous section. Total birth is estimated by constant fertility rates, that is the

assumption of constant total birth is omitted. We start our analysis from steady state,

i.e. the model is calibrated in a way that the malaria and non-malaria population grow

at the same constant rate. The interventions are applied for an indefinite period of

time, so the WHO-CHOICE methodology is abstained from. Costs are not discounted.

3.2.2 Solow model

Inspired by Solow (1956), Cuddington and Hancock (1994) and Haacker (2002a)

estimated the economic impact of HIV/AIDS in Southern Africa using a modified

Solow model from which we depart our analysis. We choose this model as it

incorporates the desired channels mentioned earlier through which malaria-

interventions affect economic well-being.

19

Using a Cobb-Douglas form, the Harrod-neutral4 production function for closed

economies in discrete time periods is the following:

αα −= 1)( itititit EAKY in which 10 << α , (1)

where itY represents income (GDP) and itA is the total factor productivity variable.

itK is the stock of physical capital and itE represents effective labour input. α is the

elasticity of output with respect to physical capital which is kept constant over time.

Subscripts i and t denote respectively region and time in years.

In the base-year, 2005, the TFP-variable, 0iA is fitted by means of equation (1). Data

for GDP, measured in constant 2000 US$, are obtained from the United Nations

Statistics division. The MNP-institute provided us with annual data on capital stock,

expressed in constant 2000 USD, and elasticities to physical capital. Labour input for

the base-year we obtained from our demographics and epidemiology model.

In our model, TFP evolves over time according to:

ititititi AgA ))(1( 1)1( −+ Ψ−Ψ++= γ (2)

where γ represents a constant universal elasticity adjusting TFP for malaria

incidence, i.e. γ is an estimation of the indirect effects of malaria on economic well-

being. To this end, we use the calculations of Cole and Neumayer (2006). The authors

argue that poor health negatively affects labour productivity and human capital

accumulation and that labour and capital are not allocated efficiently. As data

limitations deter exact examination of these linkages, poor health is best modelled

when determining its impact on TFP. They estimate the impact of malaria on TFP

using a production function composed of physical capital, human capital and total

labour stock. As an indicator of malaria, the authors use the one provided by Gallup et

al. (1999), which is the product of land area subject to malaria and the percentage of

malaria P. falciparum cases. When estimating the impact of poor health on TFP by

4 That is, technological progress is exogenous and labour augmenting.

20

means of a regression, trade openness, share of agricultural value added to GNP and

the inflation rate are included as control variables. The analysis is applied to a panel

of 6 regions using 5-yearly intervals between 1965-1995. Fixed effects are reported as

a Hausman test suggests that country effects are correlated with the explanatory

variables and therefore inconsistencies exist. Cole and Neumayer find that a 1%

increase in malaria incidence will reduce TFP by 0.58-0.75%. As we believe that the

countries used in their study are representative for our analysis, we use their

calculations when determining the economic impact of malaria interventions

indirectly, using the low impact scenario of 0.58%.

itΨ represents the proportional reduction in total malaria incidence summed over sex

and age over time and across region. So, 1−Ψ−Ψ itit is the difference in proportional

malaria incidence in two consecutive time periods. Our conservative implementation

approach for the malaria vaccine program forces us to model the indirect effects of

malaria suppression on TFP in this way.

In (2) ig is the constant non-intervention TFP-growth rate that is estimated on basis

of historical data over the period 1990-2005, by means of equations (1) and (2) with

01 =Ψ−Ψ −itit . We find TFP growth rates to range between -1.0% (Central Africa)

and + 1.1% (Eastern Africa). As pointed upon by Sorensen et al. (2005), TFP growth

rates have to be positive to reach a plausible steady state level and to accord with

empirics in the long run. Therefore, we altered Central Africa’s TFP growth rate to

ensure all focus regions to experience positive TFP growth. We assigned the lowest

TFP-growth rate across regions, i.e. that of region 1.

Next, the physical capital variable, itK , evolves over time according to:

itiititititi NcKYK Ω−−+=+ εδτ )1()1( , (3)

21

where iτ is the economy’s savings rate (MNP, 2004), which is assumed to be

constant over time5 and δ is the constant universal depreciation rate, that is set at 8%

(Haacker, 2002a)6. The last term of equation (3), iitit cNΩε , is the cost of malaria

interventions that is carried by the economy under consideration and hampers capital

accumulation. When omitting this term, one estimates the gross effects of malaria

interventions. iti Nc is the product of the constant per capita cost of the intervention

package (see appendix B) and total population. Ω is the constant universal fraction of

savings that is diverted away from capital accumulation as a result of malaria

intervention costs. We assume that two thirds of the total costs of malaria

interventions are funded externally (based on the World Health Report 2008).

Assuming that a quarter of these grants crowds out other investment projects, a

decline of 17% (i.e. 67% * 25%) in savings is experienced. The one third of total

malaria cost that is financed by the economy under consideration decreases disposable

income, we assume to an extent of 80%. As we consider a closed economy, 20% of

the total costs borne by the economy deplete capital accumulation. Then, in effect, a

6% (i.e. 33% * 20%) decline in savings occurs on the account of funding internally

the intervention costs. Summing up, we arrive at a Ω of 23%. When the economy

under consideration will bear these costs, we assume other things, such as current

government spending, being equal. Then, iit cNΩ , is the proportion of total costs of

the intervention that impedes capital accumulation, )1( +tiK . We include adjustment

factor itε as the steady state growth rates of all variables, with exception of the last

term, in equation (3) are equal to the sum of the population growth rate, in , and TFP

growth, ig . In steady state iit cNΩ grows at a lower rate, in . Then, steady state is not

attained. To adjust for this anomaly, we include itε to ensure that all variables in (3)

5 Although Nur (1993) states that households who are affected by malaria have to hire labour when incapable to work and therefore savings are decreased, in our analysis we keep saving rates constant in line with the assumptions of Solow (1956). This reasoning is ratified by Chima et al. (2002) who argue that too little studies are present to extrapolate field-studies on saving behaviour in relation to malaria to nation-wide estimates. The one that is present (Shephard, 1991) is entrenched with difficulties. Given our spectrum of countries we feel keeping saving rates constant gives a better reflection of reality. Therefore, we deviate from Haacker (2002a) who modelled total investments on the basis of household savings and HIV/AIDS prevalence. 6 This is a substantial depreciation rate. However, we feel that using this rate is justified as we focus on African countries.

22

grow at the same rate. We normalize 0 1i

ε = and assume that it grows at the TFP

growth rate.

Effective labour, itE , as a production factor of gross domestic product, encompasses

both a measurement of quantity and quality of the labour force. It is composed of the

malaria- and non-malaria population, and is defined as following:

∑∑= =

−=2

1

65

15

)(m k

itmkimkimkit LbzHSVE (4)

Total labour force, which is estimated in the demographics and epidemiology model,

across regions, i, over time, t, sex, m, and 1 year-age cohorts, k, is denoted by itmkL .

We consider the active labour force to be aged 15-65. No adjustments are made for

unemployment due to poor data availability as done in similar studies, among that of

Abegunde et al. (2007). We do correct for productivity differences across region, age

and sex by means of health state valuations (HSV). The parameter z, constant over

time and across regions, indicates the fraction of work lost per worker due to malaria

incapacitation. It is equal to one minus the universal disability weight of malaria.

Malaria prevalence across region, age and sex is captured by imkb . We deviate from

Cuddington and Hancock (1994) by disallowing work experience as a factor of labour

efficiency. In contrast to HIV/AIDS, work experience is not relevant in case of

malaria as adult malaria mortality is insignificant. Therefore, this variable, which in

the literature is modelled to be only dependent on age (see e.g. Cuddington (1993,

1994), Haacker (2002a)), will not influence economic development in case of malaria.

Rearranging equation (4) gives the following:

∑∑

∑∑

∑∑∑∑

=

−

=

=

−

=

= == =

+

=−+

=−=−=

2

1

65

15

2

1

65

15

2

1

65

15

2

1

65

15

)*_*(

)*)(*(

)**()(

m

malpop

itmkmal

malpopnon

itmkimk

k

m

malpop

itmk

malpop

itmk

malpopnon

itmkimk

k

malpop

itmk

m k

itmkimk

m k

itmkimkimkit

LweightdisablityLHSV

LzLLHSV

LzLHSVLzbHSVE

23

(5)

Equation (5) simply states that the labour population is divided into a non-malaria and

into a malaria population in which we consider the latter be discounted by the malaria

disability weight. The former is adjusted for the region-specific health state valuations

for the general population.

3.2.3 Solow’s steady state

When intervening in malaria for an indefinite period of time, the malaria and non-

malaria population growth rates, and therefore total population growth, in , become

constant over time. Therefore, the proportional difference in malaria incidence,

1−Ψ−Ψ itit , gets equal to zero as the interventions are fully implemented over time. So,

equation (2) is reduced to ititi AgA )1()1( +=+ . With the other parameters constant as

well, our model fulfils Solow’s conditions to reach steady state in which all regions

will experience balanced growth over time. The per worker GDP and capital steady

state levels are derived below.

Given the constant total population growth rate in the long run and the fact that the

effective labour population is derived from total population we use the following

characterization of the evolution of effective labour:

ititi EnE )1()1( +=+ (6)

Then the technology adjusted per capita production function, in which the lower-case

symbols with a tilde represent the per effective worker technology adjusted variables

in capital letters, itit

it

itEA

Kk =~

and itit

it

itEA

Yy =~ , is obtained by dividing (1) by the

labour force as defined in (6) and adjusting for technology, itA . This gives:

αitit ky

~~ = (7)

24

Given that ititi AgA )1()1( +=+ and ititi EnE )1()1( +=+ , and replacing the steady state

value of cN itit Ωε over Yit

by iΦ equation (3) can be rewritten as follows:

)~

)1(~)(()1)(1(

1~)1( ititii

iti

ti kygn

k δτ −+Φ−++

=+ with it iti

it

N c

Y

ε ΩΦ = (8)

Given that αitit ky

~~ = and solving (8) for itti kk~~

)1( =+ we find the steady state level for

technological adjusted capital per worker, *~itk . This gives:

)1/(1

* )(~α

δ

τ−

+++

Φ−=

iiii

ii

igngn

k (9)

By means of equations (7) and (9) we derive the technology adjusted output per

capita:

)1/(

* )(~αα

δ

τ−

+++

Φ−=

iiii

ii

itgngn

y (10)

In per capita effective terms, equations (10) and (11) become:

)1/(1

* )(α

δ

τ−

+++

Φ−=

iiii

ii

ititgngn

Ak and

)1/(

* )(αα

δ

τ−

+++

Φ−=

iiii

ii

ititgngn

Ay (11)

From (11) we can derive some important statements. Per worker capital and output

steady state levels decrease as the population growth, in , and the cost of malaria

interventions, iΦ , increase. If an intervention brings down malaria prevalence and

productivity loss due to malaria (captured by b and z respectively), then the stock of

effective labour, itE , will increase and per capita steady state levels consequently

25

decrease. Higher saving rates, iτ , and a lower depreciation rate, δ, have the opposite

effect. Moreover, lower malaria prevalence rates increase itA indirectly and therefore

as well per capita capital and output steady state levels. So, malaria-interventions

affect per capita output through their effect on iiiit bnA ,,, Φ , and z in opposite ways.

Moreover, from (11) we acknowledge that the steady state growth rates for per worker

output and capital are equal to ig . Redefining (11) in terms of total GDP and capital

steady state levels, we see that both grow at rate ii ng + .

The model as defined in equation (1) to (11) is labelled as the “closed-economy

model” and will be our model to analyze and project the economic impact of malaria-

interventions. Our analysis covers a period of 45 years and starts from steady state:

we calibrated the Solow model by means of the 2005 capital stock. Table 5 shows the

input values of the Solow model across regions in 2005.

Input variable region 1 region 2 region 3 region 4

elasticity to physical capital (α) 0.5 0.5 0.5 0.5

depreciation rate (δ) 8% 8% 8% 8%

savings rate (s) 29% 24% 12% 23%

TFP growth rate (gi) 0.5% 0.5%* 1.1% 0.8%

Effective labour stock (millions) 149 45 121 67

initial GDP per effective worker (y) 670 711 480 836

initial capital per effective worker (k) 1812 1505 485 1792

initial population growth rate (n) 2.33% 2.38% 2.45% 1.27%

Table 5 Basic parameters Solow model across region. * TFP growth rate is similar to that of region 1, the lowest rate across regions.

3.2.4 Growth accounting

In the following, we conduct GDP growth accounting by means of the production

function as defined by (1). Not only is growth accounting of importance to identify

the sources of economic growth, but also to point to the production factors through

which malaria interventions affect economic well-being.

Taking logs and time differences from (1) we get:

)ln)(ln1(

)ln(ln)ln)(ln1(lnln

)1(

)1()1()1(

itti

ittiittiitti

EE

KKAAYY

−−+

−+−−=−

+

+++

α

αα (12)

which can be written in growth rates ig as:

26

E

it

K

it

A

it

Y

it gggg )1()1( ααα −++−= (13)

We see from (13) that GDP grows at rate )( ii ng + given that K

itg grows at rate

)( ii ng + in its steady state.

3.2.5 Sensitivity analysis Solow model

By means of a Tornado analysis we tested the sensitivity of our model by varying the

values of several parameters separately and projected their impact on GDP per capita.

We used both positive and negative intervals of 5%, 10%, 15%, 20% and 25% of the

depreciation rate, elasticity to physical capital, TFP-growth- and savings rate. Other

variables were left out of the analysis as these were real observations or a fitted value

of these observations (TFP in 2005). One must keep in mind that this sensitivity

analysis scrutinizes parameters individually, not in unity. This is a weakness as

parameters may reinforce each other. However, by taking large variances this

limitation is to be compensated. We found no abnormalities when conducting the

sensitivity analysis. That is, varying the variables mentioned above substantially did

not alter model outcomes substantially.

4. Results

4.1 Cost-effectiveness study

Given our data and model assumptions, increasing coverage levels from current levels

as exposed in table 4 to 80% shows that all evaluated malaria control interventions are

highly cost-effective with averages of 8-126 int$/DALY across all regions. The

results of all interventions across regions are shown in tabular and graphic form in

appendix C.

The expansion paths indicated by the lines that connect the dots that lie closest

to the south-east corner of figures 1-4 show the order of interventions in terms of

health gains given certain budget constraints. From our analysis we conclude that this

order is the same irrespective of region, i.e. the same intervention packages are most

cost-effective in all regions. However, in general, the health impact of the

interventions over their total costs as well as their incremental cost-effectiveness is

more favorable in regions 1 and 2 than in the other regions.

27

When resources are limited, we conclude that intervening in malaria by means

of ACT case management is most cost-effective, especially in regions 1 and 2 where

cost-effectiveness ratios range between 9.22 and 7.55 int$/DALY, respectively. For

regions 3 and 4 these ratios are 21.18 and 30.42 int$/DALY, respectively.

Complementing this control intervention with a vaccination program would increase

health gains most at lowest cost, at incremental cost-effectiveness ratios between 42-

76 int$/DALY across regions. The last intervention package that lies on the expansion

paths that achieves highest gains at lowest cost is adding to the previous package ITN

and IRS. Then differential cost-effectiveness increases between 26-83 int$/DALY

across regions.

We conclude that developing a malaria vaccine is only cost-effective when it is used

in combination with at least ACT. Although a vaccine’s effectiveness is not impeded

by patient’s behavior and pharmacokinetics and therefore this intervention has a

substantial practical advantage over other malaria control interventions, the cost-

effectiveness of a vaccine is high when used alone compared to other interventions.

However, the assumption that the vaccine proportionally reduces incidence and case

fatality to the same extent as ITN and IRS (see table 3) influences the statement made

above strongly. Therefore, we believe that an analysis on the efficacy of a malaria

vaccine on case fatality is of great importance.

Based on our analysis we strongly recommend that a swift change in curative

intervention from CQ to ACT is made. When doing so, substantial higher health gains

are achieved at approximately the same costs. Even for region 4, where current ACT

coverage is relatively high, this gain is substantial. Fortunately, this policy change is

endorsed by most countries in June 2008 (Malaria report, 2008).

When financial resources are substantial, but limited, we recommend to

complement the intervention package composed of ACT and vaccine with either IRS

or ITN as these two packages lie closely to the expansion paths displayed in figures 1-

4. Then substantial gains are achieved at the expense of a small decrease in efficiency.

We believe such a sacrifice is not justifiable in relation to other interventions

packages.

Our results diverge from that of Morel et al. (2005), the only study we compare with

as Breman’s (2006) methodology is unclear and that of Goodman et al. (2000) comes

28

across with that of Morel et al. The differences in outcomes are ascribed to the

differences in epidemiological data. Korenromp (2005) estimates total malaria

incidence in 2005 to be equal to 233 million. The GBD approximates total incidence

equal to 340 million (2002). Differences in the estimations of the WHO mortality

database (2005) and GBD (2002) are smaller, the latter predicts 75.000 more malaria

fatalities. So, our results on curative interventions are approximately the same and our

cost-effectiveness results on preventative interventions are substantially lower. It

should be noted that the GBD’s age- and sex-distribution on malaria incidence and

mortality is unclear.

In table 6 the average reductions in under 5 malaria mortality as a consequence of the

three most cost-effective interventions are shown.

scenario no. scenario Region 1 Region 2 Region 3 Region 4

5 ACT 31% 33% 25% 16%

11 Vacc + ACT 39% 41% 34% 27%

19 ITN + IRS + Vacc + ACT 47% 49% 44% 38%

Table 6 Average reductions in under 5 mortality as a consequence of certain malaria interventions.

We see that substantial reductions in under 5 mortality are achieved when intervening

by means of the most cost-effective interventions. Again the impact is larger across

intervention packages for regions1 and 2 in which reductions of 31-49% are obtained.

For regions 3 and 4 the mortality impact ranges between 16-44%. The intervention

package composed of all three preventative interventions complemented with ACT

yields the highest gain in terms of malaria mortality under 5’s. Therefore we conclude

that when intervening with this package, MDG’s 4 and 6 are best served.

Eliminating under 5 and total malaria mortality by means of the interventions

packages as defined in table 4 is unattainable. Even increasing coverage levels to

100% is insufficient to impede malaria mortality. Then, at most reductions in malaria

mortality up to 85% are attainable. So, the net effectiveness of individual

interventions have to increase to eradicate mortality as a consequence of malaria.

Given the baseline efficacy of malaria interventions as shown by Morel et al. (2005)

we believe substantial gains in effectiveness are attainable on the account of patient’s

behavior, pharmacokinetics and biogenetics.

29

4.2 Macroeconomic analysis

We conclude from our analysis that the total costs that refrain capital from

accumulating fully are negligible in a macroeconomic context. This is understandable

as the proportion of the costs of the most expensive intervention package discounted

for the fraction of savings that is diverted away from capital accumulation over total

GDP ranges between 0.01-0.03% across regions. So, contrary to a cost-effectiveness

point of view, macroeconomists do not take into account interventions costs when

deciding on intervening in malaria. This is in line with Barlow’s (1967) conclusion.

After the intervention packages are installed, the new steady state is reached after

approximately 120 years. This is understandable as is takes time before the impact of

the control methodologies are seeped through the population and their children. Also,

the Solow model has to adjust itself to these changes in effective labor. According to

Solow’s 1956 theory, in this new steady state, total capital stock and total GDP grow

at a new rate, that is equal new population growth rates, erventionpostii ng int)( −+ . This is

equal to the sum of the pre-intervention growth rates and the increase in population

growth rates, which equals ierventionpreii nng ∆++ −int)( .

It is important to understand the dynamics of the effective labour variable.

Curative interventions reduce mortality and therefore fuel effective labour growth. In

the unlikely event that only malaria incidence is reduced but mortality stays constant,

an increase in the effective labour stock is experienced due to productivity gains: the

more productive non-malaria population becomes larger and the less efficient malaria

population stock decreases. Yet, in the end, effective labour growth stays the same as

mortality stays constant. However, preventative interventions also negatively affect

malaria mortality.

The dynamics described above are clearly shown in graphs 5 and 6 which

show the influence of a preventative (ITN) and a curative (ACT) intervention on the

effective labour growth rates, respectively. The slight increase in effective labour

growth rate displayed in figure 5 is attributable to the decline in malaria mortality

among adults. However, after a few years, the growth diminishes slightly as the

labour force that survived malaria may die of other causes, i.e. total background

mortality increases. The under 5’s that that did not die of the consequences of malaria

due to the ACT-intervention enter the labour force in 2015 onwards. This is seen in

30

the increase in the effective labour force growth rate. From 2030 forwards we see

again decreased effective labour growth due to increased background mortality. This

loss is recuperated in 2038 onwards as fertility has increased due to increased

population stock.

The growth path of effective labour when intervening preventatively by means

of ITN is displayed in graph 6. The sharp rise in effective labour growth rate in the

year of intervention is attributed to the productivity gain that is attained as more

people become healthy, i.e. the proportion of the malaria-population over the non-

malaria population becomes smaller. This is confirmed by the fact that malaria

population growth rates fall sharply after implementation, the non-malaria population

grows harshly. However, this augment evaporates, be it partially as curative

interventions also negatively affect malaria mortality. After this, the dynamics of the

growth path are the same as when one intervenes curatively.

Combinations of curative and preventative interventions inhibit growth paths

characteristics of both types.

The increases in total capital’s and GDP’s steady state growth rates are equal to the

increases in effective labour growth rates. Therefore our results accord with Solow’s

theory. Table 10 in appendix D shows the results on GDP steady state growth rates

across regions and interventions. Steady state GDP per capita growth across region is

equal to ig irrespective of the intervention.

We conclude from tables 8 and 10 that in general those interventions that have

the highest gain in terms of DALY’s averted, achieve the highest effective labour

growth rates and therefore stimulate GDP steady state growth most. This makes sense.

When calculating the gains in years lived with disability (YLD), a measurement of

morbidity and therefore a measurement of productivity, over the gains in DALY’s, we

see that this proportion is large just after the interventions are implemented. After a

short period of time, however, the importance of morbidity in total health gains

decreases substantially implying that averted mortality plays a more important role

over time. Subsequently, proportionally averted mortality contributes more to

DALY’s than morbidity does. Due to decreased mortality, effective labour growth

increases. Then, the relationship between DALY’s and steady state GDP growth is

established and therefore effectiveness gains in terms of DALY’s is what counts given

31

that mortality contributes more to overall health gains than morbidity does. So,

intervening by means of ITN, IRS, ACT and a vaccination is not only cost-effective in

terms of int$/DALY, but also stimulates GDP steady state growth most in all regions.

It is of importance to note that in the new steady state, the capital-output ratio,

ii YK / , has decreased. This implies that the importance of capital as a contributor of

economic growth has diminished over time. This makes sense. When dividing the

steady state levels of capital over output, we see that an increase in the population

growth rate decreases the steady state capital-output ratio.

4.2.1 Direct effects of malaria interventions

The moment the interventions are implemented, the demographics and epidemiology

and therefore the Solow model get out of balance. As theory predicts, during the

transition to a new steady state, GDP growth accords to the growth accounting

exercise given by equation (13).

As TFP-growth is constant, GDP growth is only influenced by the growth

differentials of effective labour and total capital stock, that are multiplied by )1( α−

and α , respectively. From the total capital accumulation function (3), we see that

differences in total capital are influenced by total GDP and total capital in the

previous year. However, their influences are discounted for by the savings and

depreciation rate. So, when the malaria control interventions affect effective labour

growth, GDP growth is positively influenced. Consequently, total capital

accumulation and therefore capital growth is influenced, however to a lesser scale due

to the damped influences of the depreciation and savings rate on GDP and total

capital, respectively. These dynamics are clearly shown in graphs 5 and 6 where

capital growth rates do not follow the same bumpy track as GDP growth rates. Instead

over time capital growth increases more smoothly.

We evaluate the direct impact of the malaria interventions on GDP per capita in 2070

and express the results as a proportional difference compared with GDP per capita of

the non-intervention scenario. See table 11, appendix D.

From the results in table 11 emerges the same picture as for total GDP growth

rates, however in inverse: in general, those interventions that achieve the highest gain

in terms of DALY’s affect GDP per capita in 2070 worst. We find that the

interventions directly decrease GDP per capita levels between 0.4-1.8% in 2070

32

across regions. This is understandable, the pie of GDP gets larger, however it has to

be divided by more inhabitants. We see that population grows faster than GDP, so

GDP per capita is smaller when intervening in malaria compared to a non-intervention

scenario. So, where population has grown fastest, and therefore where the health gains

in terms of DALY’s are greatest, GDP per capita is worst affected.

4.2.2 Direct effects complemented with indirect effects

The indirect effects are modelled through the inverse relationship between malaria

incidence and TFP growth. An instant increase in TFP occurs in all preventative

interventions, with the exception of a malaria vaccine. As a conservative approach for

implementing this medication is followed, malaria incidence decreases gradually over

time. Figure 7 and 8 show the reduction in malaria incidence and growth paths of the

TFP, total capital and GDP for ITN and the malaria vaccine respectively over time for

region 1.

From both graphs we see that the mechanisms through which the malaria

interventions directly affect GDP are overshadowed by TFP growth which is driven

by the decrease of malaria incidence. From equation (13) we see that the impact of

TFP growth on GDP growth is multiplied by )1( α− . Again, from definition (3), we

see the impact of TFP on total capital growth is mitigated and delayed. In steady state,

TFP growth rate returns to its pre-intervention steady state level.

The main difference between intervening by means of ITN or a vaccine is the

time-frame in which malaria incidence is decreased. The former achieves an instant

decrease as we assume that coverage of the malaria interventions are directly

implemented, see graph 7. Therefore a one-time peak in TFP growth is realised. As

we assume a gradual implementation of the malaria vaccine, a more bumpy course of

TFP growth arises, see figure 8.

We are aware that a large increase in TFP growth is unrealistic, however we

assess the results in terms of GDP per capita after a time span of 65 years. So,

whether interventions are gradually or instantly implemented, we argue that in the end

our results are not altered by the course of TFP growth.

As TFP drives GDP growth to the largest extent and the fact that TFP is

modelled by means of incidence reduction, it is no surprise that those interventions

that yield the highest gain in proportional GDP per capita are the preventative

interventions. The combination of ITN, IRS and a vaccine result in an increase in

33

GDP per capita between 43.5-49% in Africa in 2070. The least effective preventative

intervention still yields a proportional increase in GDP per capita of 14.2% (see table

11 for an overview of all results).

The malaria control interventions should be scaled up as the economic gains as

a consequence of malaria suppression are substantial. However, as malaria mortality

and TFP are not related, our results show that solely intervening by means of ACT or

CQ do not induce the complementary gains that preventative interventions do yield.

Therefore, the most cost-effective interventions with the exception of ACT stimulate

economic development in terms of GDP per capita. Though, intervening through

ACT, ITN, IRS and a vaccine yield the highest gain in GDP per capita.

Overall, our analysis shows that the development of a vaccine is cost-effective when

used in combination with other intervention methodologies, either ACT or ACT in

combination with the other preventative interventions, IRS and ITN. More

importantly, intervening by these means augments economic development

substantially. Our study shows that intervening by means of ACT, ITN, IRS and a

vaccine is cost-effective, but yields sub-optimal results in terms of macroeconomic

gains.

5. Discussion

All studies have their assumptions that influence outcomes to a certain extent, this one

included. Many debate the effectiveness of the malaria control interventions

considered in this study. Morel et al. (2005) provide a nice overview of this

discussion. Moreover, our assumption that a malaria vaccine reduces case fatality to

the same extent as ITN and IRS is of importance when deciding on the

implementation of a malaria vaccine. That is, the cost-effectiveness of the vaccine

differs substantially if this medication reduces case fatality to a larger or smaller

extent than assumed. Therefore, we believe more research on this topic should take

place. However, one should take into account that, when immunized, behavioural

characteristics of the patient and pharmacokinetics do not influence the effectiveness

of the vaccine. We believe this to be a great advantage with respect to the other

intervention methods. In addition, we feel that policy makers should take into account

local influences when deciding on the malaria control interventions. The cost-

34

effectiveness study is not a blue-print for reducing the malaria burden given certain

budget constraints.

When determining the macroeconomic impact of the malaria interventions several

assumptions are debatable. Firstly, we implicitly assume that the effectiveness of the

interventions remains constant during our period of analysis. Given that resistance to

anti-malaria medication is likely to occur over time, this is a crude assumption.

Secondly, Cole and Neumayer (2006) assume linearity in the relationship between

TFP and malaria incidence. It is questionable to what extent such a relationship can be

maintained. Moreover, when estimating the indirect effects of malaria on economic

well-being, we ask ourselves whether this should be related to incidence.

The Solow model has some drawbacks. Mills and Shillcutt (2004) argue that

capital accumulation in Solow models is overemphasized as a production factor of

economic growth. Secondly, Sorensen et al. (2005) point to a limitation articulated in

the assumption of the exogenous technological growth rate. Thirdly, Solow assumes

the savings rate to be exogenous and constant over time. Ros (2003) reasons that this

rate can rise as a consequence of increasing income levels or increasing marginal

returns on capital. Considering the last two critiques, Solow models do not inhibit any

feedback mechanisms and therefore we argue that these models are quite static.

Given these drawbacks and assumptions that strongly influence our study, we

nevertheless believe that our methodology is suitable for simulating the economic

impact of malaria interventions for several reasons. Firstly, the Solow model is the

most commonly used workhorse model to discuss economic growth due to its

simplicity. It estimates economic well-being in the long run by taking into account

key macro-economic aggregates which are also commonly referred to in context of

malaria. Secondly, considering the scarcity and the poor quality of data in countries

where malaria is prevalent, the Solow model is still able to articulate on the course of

an economy. Thirdly, the Solow model used in this study allows us to identify the

direct and indirect effects of malaria interventions on economic growth through its

impact on TFP. For these reasons, we consider the Solow model the tool of choice to

organize thinking about the macro effects of malaria interventions.

35

6. Conclusion

Our study shows that in sub-Saharan Africa the development of a malaria vaccine is

only cost-effective when this malaria control intervention is used either in

combination with ACT or when it is complemented with ACT, ITN and IRS. The

latter intervention package serves the Millennium Development Goals best. We found

that intervening solely by means of ACT is also cost-effective. Moreover, this study

concludes that elimination of malaria mortality is not feasible given the current

effectiveness of the interventions and combinations thereof.

Intervening in malaria stimulates economic development substantially, given

that curative interventions are not used solely but in combination with preventative

interventions. Our cost-effectiveness and macroeconomic impact studies do not find a

common intervention package that yield optimal results.

36

References

Abegunde, D.O., Mathers, C.D., Adam, T., Ortegon, M, Strong, K., 2007. The burden and costs of chronic diseases in low-income and middle-income countries. Lancet 370;1929-38. Aponte,J.J., Aide,P., Renom,M., Mandomando,I., Bassat,Q., Sacarlal, J., Manaca, M.N., Lafuente, S., Barbosa,A., Leach,A., Lievens, M., Vekemans, J., Sigauque, B., Dubois, M.C., Demoitié, M.A., Sillman,M., Savarese, B., McNeil, J.G., Macete, E., Ballou, W.R., Cohen, J., Alons, P.A., 2007. Safety of the RTS,S/AS02D candidate malaria vaccine in infants living in a highly endemic area of Mozambique: a double blind randomised controlled phase I/IIb trial, Lance;, 370: 1543–51.

Arndt, C., Lewis, J.D., 2001. The HIV/AIDS Pandemic in South Africa: Sectoral Impacts and Unemployment. Journal of International Development, Vol. 13 (May), pp. 427-49. Baltussen, R., Tan-Torres, E.T., 2003. Methods for generalised cost-effectiveness analysis: a guide. WHO guidelines on cost-effectiveness studies. WHO, Geneva.

Barlow, R., 1967. The economic effects of malaria eradication. American EconomicReview 57(2): 130-48. Barro, R. J., 1991. Economic growth in a cross-section of countries. Quarterly Journal of Economics 106: 407-43. Bloom, D.E., Canning, D., Sevilla, J., 2004. The effect of health on economic growth: a production function approach. World Development Vol.32, No.1, 1-13. Bloom, D. E., Mahal, A.S., 1997. Does the AIDS Epidemic Threaten EconomicGrowth?. Journal of Econometrics, Vol. 77 (March),105-24. Bonnel, R., 2000. HIV/AIDS: Does it Increase or Decrease Growth in Africa. Unpublished paper, Washington: World Bank. Borts, G.H. 1967. Discussion of “The economic effects of malaria eradication” by Barlow (1967). The American Economic Review, Vol. 57, No.2. Breman, J. G. Alilio, ,M. S.,Mills, A., 2004. Conquering the Intolerable Burden of Malaria: What’s New, What’s Needed: A Summary. American Journal of Tropical Medicine and Hygiene 71 (Suppl. 2): 1–15.

Breman, J.G., Mills, A., Snow, R.W., Mulligan, J.A., Lengeler, C., Mendis, K., Sharp, B., Morel, C., Marchesini, P., White, N.J., Steketee, R.W., Doumbo, O.K., 2006. Disease control priorities in developing countries, 2nd edition, New York, Oxford University press. Brown, P.J., 1986. Socioeconomic and demographic effects of malaria eradication: a comparison of Sri Lank and Sardinia. Soc. Sci. Med., Vol. 22, No.8, pp. 847-859.

37

Bryce, J., Boschi-Pinto, J.B., Shibuya, K., Black, R.E., and the WHO child mortality reference group, 2005. WHO estimates of the causes of deaths of children. Lancet 365 (9464): 1114-6. Chima, R. I., Goodman, C. A., Mills, A., 2003. The economic impact of malaria in Africa: a critical review of the evidence. Health Policy 63: 17-36. Cole, M. A., Neumayer, E., 2006. The impact of poor health on total factor productivity. Journal of Development Studies, 42:6, 918— 938 Cuddington, J.T., J.D. Hancock, 1992. Assessing the impact of AIDS on the growth path of the Malawian economy, Working paper $92-07 (Georgetown University, Washington, DC). Cuddington, J.T., Hancock, J.D., 1994. Assessing the impact of AIDS on the growth path of the Malawian economy. Journal of Development economics 43, 363-368. Ettling, M.B., Shephard, D.S., 1991. Economic cost of malaria in Rwanda. Tropical Medicine and Parasitology, vol. 42, no. 3, pp. 214-8. Frederiken, H., 1966. Malaria eradication and population growth. Am. J. trop. Med. Hyg., 15, 261. Gallup, J.L., Sachs, J.D., 1998. The economic burden of malaria. Harvard Center for International Development. Gallup, J.L, Sachs, J.D., Mellinger, A.D., 1999. Geography and economic development, International Regional Science Review; 22; 179.

Goodman, C., Coleman, P., Mills, A., 2000. Economic analysis of malaria control in sub-Saharan Africa. Global forum for health research. Haacker, M., 2002a. Modelling the macroeconomic impact of HIV/AIDS. IMF working paper, WP/02/195. Haacker, M., 2002b, The Economic Consequences of HIV/AIDS in Southern Africa, IMF working paper 02/38. Korenromp, E., 2005. Malaria incidence estimates at country level for the year 2004. RBM Monitoring and Evaluation Reference Group & MERG Task Force on Malaria Morbidity World Health Organization, Roll Back Malaria.

Leighton, C., Foster, R., 1993. Economic impacts of malaria in Kenya and Nigeria. Bethesda, Maryland: Abt Associates, Health Financing and Sustainability Project.. Cited in: Goodman et al. (2000). MacFarlan, M., Sgherri, S. 2001. The Macroeconomic Impact of HIV/AIDS in Botswana. IMF working paper, WP/01/80.

38

McCarthy, F.D., Wolf, H., Yi, W., 2000. The growth costs of malaria. National bureau of economic research. Working paper 7541. Mills, A., Shillcutt, S., 2004. Challenge Paper on communicable diseases. Copenhagen Consensus. Murray, C. J. L., Lopez, A. D., 1990. The Global Burden of Disease: Volume 1. Geneva: World Health Organization, Harvard School of Public Health and The World Bank.

Murray C.J.L., Lopez A.D., 1996. The global burden of disease: A comprehensive assessment of mortality and disability from diseases, injuries and risk factors in 1990 and projected to 2020. Harvard School of Public Health (on behalf of WHO and the World Bank), distributed by Harvard University Press. Nájera J.A., Hempel J., 1996. The burden of malaria. WHO CTD/MAL/96.10. Niessen, L., Ten Hove, A., Hilderink, H., Weber, M., Mulholland, K., Ezzati, M., 2009. Comparative assessment of child pneumonia interventions, WHO bulletin, in press.

Nur, E.T.M., 1993. Impact malaria on labour use and efficiency in Sudan, Sot. Sci. Med. Vol. 37, No. 9, pp. 1115-1I 19.

Over, M., 1992. The Macroeconomic Impact of AIDS in Sub-Saharan Africa. The World Bank, Technical working paper, n°3. Shephard, D.S., Ettling, M.B., Brinkman, U., Sauerborn, R., 1991. The economic cost of malaria in Africa. Tropical medicine and parasitology, vol. 42, pp 197-223. Sinha,A., Levine, O., Knoll, M.O., Muhib,F., Lieu, T.A., 2007. Cost-effectiveness of pneumococcal conjugate vaccination in the prevention of child mortality: an international economic analysis, Lancet; 369: 389–96.

Solow, R. M.,1956. A Contribution to the Theory of Economic Growth. Quarterly Journal of Economics, LXX,65-94. Sørensen, P.B., Whittha-Jacobsen, H.J., 2005. Introducing advanced macroeconomics: growth and business cycles. McGraw-Hill, Berkshire.

Wang'ombe, J.K., Mwabu, G.M., 1993. Agricultural land use patterns and malaria conditions in Kenya. Social Science and Medicine, 37(9), 1121-30. Cited in: Goodman et al. (2000). WHO, 2002. World Health Report 2002: Reducing Risks, Promoting Healthy Life. Geneva: WHO.

WHO mortality database, 2005. http://www.who.int/healthinfo/morttables/en/index.html World malaria report, 2005. Roll back malaria, WHO, UNICEF. World malaria report, 2008. WHO.

39

Appendix A Gradual approach cost-effectiveness assessment

1. Construction epidemiologic and demographic model Model describes interaction demographics and epidemiology

2. Data collection Collection of the data of relevance for the impact assessment

3. Construction of baseline scenario To calculate the cost-effectiveness of the interventions a benchmark of no

intervention is established.

4. Estimation effectiveness interventions The effectiveness of the interventions that affect the epidemiological

parameters will be estimated in terms of health gains, in our case DALY’s

averted.

5. Calculation of the total costs of the interventions 6. Overview cost-effectiveness interventions

40

Appendix B

Region

scenario no. scenario Region 1 Region 2 Region 3 Region 4

1 ITN 0.76 0.66 0.64 0.63

2 IRS 0.80 0.65 0.62 0.60

3 Vacc 10.97* 10.97* 10.97* 10.97*

4 CQ 0.24 0.21 0.20 0.20

5 ACT 0.25 0.21 0.21 0.20

6 ITN+CQ 0.99 0.86 0.84 0.82

7 ITN+ACT 0.99 0.86 0.84 0.82

8 IRS+CQ 0.93 0.79 0.76 0.74

9 IRS + ACT 0.93 0.79 0.77 0.75

10 Vacc + CQ 0.24 0.21 0.20 0.20

11 Vacc + ACT 0.25 0.21 0.21 0.20

12 ITN + IRS 1.29 1.11 1.08 1.05

13 ITN + Vacc 0.75 0.66 0.64 0.63

14 IRS + Vacc 0.80 0.65 0.62 0.60

15 ITN + IRS + Vacc 1.29 1.11 1.08 1.05

16 ITN + IRS + ACT 1.36 1.17 1.14 1.11

17 ITN + Vacc + ACT 0.99 0.86 0.84 0.82

18 IRS + Vacc + ACT 0.93 0.79 0.77 0.75

19 ITN + IRS + Vacc + ACT 1.36 1.17 1.14 1.11

Table 7 Analyzed interventions and their costs (int$ per capita) across region. * per person vaccinated

41

Appendix C

Average yearly cost (int$) Average yearly effectiveness (DALY's averted)

Region Region

scenario no. scenario 1 2 3 4 1 2 3 4

1 ITN 185 49 129 66 3 1 1 1

2 IRS 195 48 124 63 3 1 1 1

3 Vacc 95 29 77 39 3 1 1 1

4 CQ 58 16 40 21 2 0 1 0

5 ACT 61 16 42 21 6 2 2 1

6 ITN+CQ 241 64 169 86 4 1 2 1

7 ITN+ACT 242 64 169 86 8 3 3 1

8 IRS+CQ 227 58 153 78 4 1 2 1

9 IRS + ACT 227 59 155 79 8 3 3 1

10 Vacc + CQ 153 45 117 60 5 1 2 1

11 Vacc + ACT 156 45 119 60 9 3 3 1

12 ITN + IRS 314 82 217 110 5 1 2 1

13 ITN + Vacc 277 78 206 105 5 2 2 1

14 IRS + Vacc 290 77 202 102 6 2 2 1

15 ITN + IRS + Vacc 409 112 294 149 7 2 3 2

16 ITN + IRS + ACT 332 87 229 117 9 3 3 2

17 ITN + Vacc + ACT 337 93 246 125 10 3 4 2

18 IRS + Vacc + ACT 322 88 232 117 10 3 4 2

19 ITN + IRS + Vacc + ACT 427 116 306 155 11 3 4 2

Table 8 Average yearly cost and average yearly effectiveness across scenario and region (in millions).

42

Region

1 2 3 4

scenario no. scenario CE ∆CE CE ∆CE CE ∆CE CE ∆CE

1 ITN 72.91 dominated 63.05 dominated 125.75 dominated 107.25 dominated

2 IRS 74.49 dominated 61.38 dominated 124.08 dominated 109.54 dominated

3 Vacc 28.82 dominated 29.62 dominated 56.93 dominated 48.26 dominated

4 CQ 38.16 dominated 39.64 dominated 43.82 dominated 62.69 dominated

5 ACT 9.42 9.42 7.55 7.55 21.18 21.18 30.42 30.42

6 ITN+CQ 62.87 dominated 57.61 dominated 93.79 dominated 95.62 dominated

7 ITN+ACT 30.08 dominated 25.35 dominated 62.27 dominated 71.65 dominated

8 IRS+CQ 59.06 dominated 52.92 dominated 84.86 dominated 86.29 dominated

9 IRS + ACT 28.09 dominated 23.24 dominated 57.36 dominated 67.48 dominated

10 Vacc + CQ 33.57 dominated 33.97 dominated 55.53 dominated 55.29 dominated

11 Vacc + ACT 17.95 42.92 16.45 43.90 39.75 76.54 43.43 56.56

12 ITN + IRS 66.59 dominated 57.76 dominated 119.14 dominated 102.05 dominated

13 ITN + Vacc 50.79 dominated 47.18 dominated 94.24 dominated 79.11 dominated

14 IRS + Vacc 52.39 dominated 46.52 dominated 93.04 dominated 78.82 dominated

15 ITN + IRS + Vacc 56.01 dominated 50.55 dominated 103.83 dominated 87.01 dominated

16 ITN + IRS + ACT 35.49 dominated 30.03 dominated 70.36 dominated 73.69 dominated

17 ITN + Vacc + ACT 33.52 dominated 29.80 dominated 68.75 dominated 68.94 dominated

18 IRS + Vacc + ACT 31.93 dominated 28.10 dominated 65.01 dominated 65.95 dominated

19 ITN + IRS + Vacc + ACT 38.25 26.23 33.74 21.28 76.23 54.84 72.98 62.88

Table 9 Average cost-effectiveness (CE) and incremental cost-effectiveness (∆CE) (int$ per DALY averted) across scenario and region.

43

ITNIRS

Vacc

CQ ACT

ITN+CQ ITN+ACT

IRS+CQ IRS + ACT

Vacc + CQ Vacc + ACT

ITN + IRS

ITN + VaccIRS + Vacc

ITN + IRS + Vacc

ITN + IRS + ACTITN + Vacc + ACT

IRS + Vacc + ACT

ITN + IRS + Vacc + ACT

0

50

100

150

200

250

300

350

400

450

0 2 4 6 8 10 12

DALY's averted (average yearly)

Co

st

(avera

ge y

earl

y)

Figure 1 Cost-effectiveness pane region 1 (in millions).

44

ITN + IRS + Vacc + ACT

IRS + Vacc + ACT

ITN + Vacc + ACT

ITN + IRS + ACT

ITN + IRS + Vacc

IRS + VaccITN + VaccITN + IRS

Vacc + ACTVacc + CQ

IRS + ACTIRS+CQ

ITN+ACTITN+CQ

ACTCQ

Vacc

IRSITN

0

20

40

60

80

100

120

140

0 1 1 2 2 3 3 4 4

DALY's averted (yearly average)

Co

st

(yearl

y a

ve

rag

e)

Figure 2 Cost-effectiveness pane region 2 (in millions).

45

ITN + IRS + Vacc + ACT

IRS + Vacc + ACT

ITN + Vacc + ACT

ITN + IRS + ACT

ITN + IRS + Vacc

IRS + VaccITN + Vacc

ITN + IRS

Vacc + ACTVacc + CQ

IRS + ACTIRS+CQ

ITN+ACTITN+CQ

ACTCQ

Vacc

IRSITN

0

50

100

150

200

250

300

350

0 1 1 2 2 3 3 4 4 5

DALY's averted (yearly average)

Co

st

(ye

arl

y a

ve

rag

e)

Figure 3 Cost-effectiveness pane region 3 (in millions).

46

ITNIRS

Vacc

CQ ACT

ITN+CQ ITN+ACT

IRS+CQ IRS + ACT

Vacc + CQ Vacc + ACT

ITN + IRSITN + Vacc

IRS + Vacc

ITN + IRS + Vacc

ITN + IRS + ACT

ITN + Vacc + ACT

IRS + Vacc + ACT

ITN + IRS + Vacc + ACT

0

20

40

60

80

100

120

140

160

180

0 1 1 2 2 3

DALY's averted (yearly average)

Co

st

(ye

arl

y a

vera

ge

)

Figure 4 Cost-effectiveness pane region 4 (in millions).

47

Appendix D

Region

scenario no. scenario 1 2 3 4

1 ITN 0.0275% 0.0262% 0.0115% 0.0151%

2 IRS 0.0283% 0.0265% 0.0113% 0.0141%

3 Vacc 0.0375% 0.0349% 0.0157% 0.0206%

4 CQ 0.0198% 0.0159% 0.0132% 0.0115%

5 ACT 0.0833% 0.0834% 0.0287% 0.0236%

6 ITN+CQ 0.0442% 0.0396% 0.0227% 0.0248%

7 ITN+ACT 0.0979% 0.0964% 0.0358% 0.0350%

8 IRS+CQ 0.0450% 0.0399% 0.0225% 0.0239%

9 IRS + ACT 0.0984% 0.0966% 0.0356% 0.0343%

10 Vacc + CQ 0.0531% 0.0475% 0.0261% 0.0297%

11 Vacc + ACT 0.1032% 0.1008% 0.0384% 0.0393%

12 ITN + IRS 0.0514% 0.0485% 0.0210% 0.0270%

13 ITN + Vacc 0.0592% 0.0556% 0.0248% 0.0325%

14 IRS + Vacc 0.0598% 0.0559% 0.0246% 0.0317%

15 ITN + IRS + Vacc 0.0781% 0.0733% 0.0323% 0.0419%

16 ITN + IRS + ACT 0.1107% 0.1076% 0.0416% 0.0441%

17 ITN + Vacc + ACT 0.1148% 0.1111% 0.0439% 0.0483%

18 IRS + Vacc + ACT 0.1151% 0.1112% 0.0438% 0.0477%

19 ITN + IRS + Vacc + ACT 0.1248% 0.1199% 0.0485% 0.0554%

Table 10 GDP's and total capital steady state growth differential accros region and intervention.

48

1.80%

2.00%

2.20%

2.40%

2.60%

2.80%

3.00%

2002

2004

2006

2008

2010

2012

2014

2016

2018

2020

2022

2024

2026

2028

2030

2032

2034

2036

2038

2040

2042

2044

2046

2048

2050

Effective labour

Capital

GDP

Figure 5 Transition paths of total GDP, capital and effective labour when intervening with ACT for region 1 over time.

49

1.80%

2.00%

2.20%

2.40%

2.60%

2.80%

3.00%

2002

2004

2006

2008

2010

2012

2014

2016

2018

2020

2022

2024

2026

2028

2030

2032

2034

2036

2038

2040

2042

2044

2046

2048

2050

Capital

GDP

Effective labour

Figure 6 Transition paths of total GDP, capital and effective labour when directly intervening with ITN for region 1 over time.

50

Direct effects Direct and indirect effects

Region Region

scenario no. scenario 1 2 3 4 1 2 3 4

1 ITN -0.36% -0.33% -0.13% -0.15% 15.4% 16.7% 21.6% 18.3%

2 IRS -0.37% -0.33% -0.13% -0.14% 15.9% 16.9% 21.2% 17.0%

3 Vacc -0.49% -0.43% -0.18% -0.21% 23.8% 25.3% 32.8% 28.2%

4 CQ -0.30% -0.23% -0.20% -0.18% -0.30% -0.23% -0.20% -0.18%

5 ACT -1.26% -1.22% -0.43% -0.36% -1.26% -1.22% -0.43% -0.36%

6 ITN+CQ -0.61% -0.52% -0.30% -0.30% 15.1% 16.4% 21.4% 18.2%

7 ITN+ACT -1.42% -1.35% -0.49% -0.46% 14.2% 15.5% 21.2% 18.0%

8 IRS+CQ -0.62% -0.53% -0.29% -0.29% 15.7% 16.7% 21.0% 16.9%

9 IRS + ACT -1.43% -1.35% -0.49% -0.45% 14.7% 15.7% 20.8% 16.7%

10 Vacc + CQ -0.73% -0.61% -0.33% -0.35% 23.5% 25.0% 32.6% 28.0%

11 Vacc + ACT -1.49% -1.39% -0.52% -0.49% 22.5% 24.0% 32.4% 27.8%

12 ITN + IRS -0.68% -0.62% -0.24% -0.29% 27.5% 29.2% 34.6% 30.3%

13 ITN + Vacc -0.79% -0.70% -0.29% -0.35% 35.6% 37.6% 44.3% 40.4%

14 IRS + Vacc -0.79% -0.70% -0.29% -0.34% 36.0% 37.8% 44.1% 39.6%

15 ITN + IRS + Vacc -1.05% -0.94% -0.39% -0.47% 43.5% 45.1% 49.5% 46.6%

16 ITN + IRS + ACT -1.58% -1.48% -0.55% -0.55% 26.4% 28.1% 34.2% 29.9%

17 ITN + Vacc + ACT -1.63% -1.51% -0.58% -0.59% 34.5% 36.5% 43.9% 40.1%

18 IRS + Vacc + ACT -1.63% -1.52% -0.58% -0.59% 34.9% 36.6% 43.7% 39.3%

19 ITN + IRS + Vacc + ACT -1.76% -1.62% -0.63% -0.68% 42.5% 44.1% 49.1% 46.3%

Table 11 Proportional differences in GDP per capita, direct effects and direct and indirect effects, across region and intervention.

51

0%

5%

10%

15%

20%

25%

30%

2006

2011

2016

2021

2026

2031

2036

2041

2046

2051

2056

2061

2066

2071

2076

2081

2086

2091

2096

2101

2106

2111

2116

2121

2126

2131

2136

2141

2146

Incidence reduction

GDP

TFP

Capital

Figure 7 Transition paths of total GDP, capital and TFP when intervening with ITN for region 1 over time.

52

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

35.0%

40.0%

45.0%

2006

2011

2016

2021

2026

2031

2036

2041

2046

2051

2056

2061

2066

2071

2076

2081

2086

2091

2096

2101

2106

2111

2116

2121

2126

2131

2136

2141

2146

TFP

Incidence reduction

Capital

GDP

Figure 8 Transition paths of total GDP, capital and TFP when intervening with a vaccine for region 1 over time.

53